arXiv subjects

Tian Wang

Publications and source records attributed to Tian Wang.

121 records · Page 7Linked to original sources

arXiv2017

SAM: Semantic Attribute Modulation for Language Modeling and Style Variation

This paper presents a Semantic Attribute Modulation (SAM) for language modeling and style variation. The semantic attribute modulation includes various document attributes, such as titles, authors, and document categories. We consider two types of attributes, (title attributes and category attributes), and a flexible attribute selection scheme by automatically scoring them via an attribute attention mechanism. The semantic attributes are embedded into the hidden semantic space as the generation inputs. With the attributes properly harnessed, our proposed SAM can generate interpretable texts with regard to the input attributes. Qualitative analysis, including word semantic analysis and attention values, shows the interpretability of SAM. On several typical text datasets, we empirically demonstrate the superiority of the Semantic Attribute Modulated language model with different combinations of document attributes. Moreover, we present a style variation for the lyric generation using SAM, which shows a strong connection between the style variation and the semantic attributes.

cs.CL↗

arXiv2017

Scaling of Majorana Zero-Bias Conductance Peaks

We report an experimental study of the scaling of zero-bias conductance peaks compatible with Majorana zero modes as a function of magnetic field, tunnel coupling, and temperature in one-dimensional structures fabricated from an epitaxial semiconductor-superconductor heterostructure. Results are consistent with theory, including a peak conductance that is proportional to tunnel coupling, saturates at $2e^2/h$, decreases as expected with field-dependent gap, and collapses onto a simple scaling function in the dimensionless ratio of temperature and tunnel coupling.

cond-mat.mes-hall↗

arXiv2017

Disentangling intra-cycle interferences in photoelectron momentum distributions using orthogonal two-colour laser fields

We use orthogonally polarized two-colour (OTC) laser pulses to separate quantum paths in multiphoton ionization of Ar atoms. Our OTC pulses consist of 400~nm and 800~nm light at a relative intensity ratio of 10:1. We find a hitherto unobserved interference in the photoelectron momentum distribution, which exhibits a strong dependence on the relative phase of the OTC pulse. Analysis of model calculations reveal that the interference is caused by quantum pathways from non-adjacent quarter cycles.

physics.atom-ph↗

arXiv2016

Non-invasive detection of animal nerve impulses with an atomic magnetometer operating near quantum limited sensitivity

Magnetic fields generated by human and animal organs, such as the heart, brain and nervous system carry information useful for biological and medical purposes. These magnetic fields are most commonly detected using cryogenically-cooled superconducting magnetometers. Here we present the frst detection of action potentials from an animal nerve using an optical atomic magnetometer. Using an optimal design we are able to achieve the sensitivity dominated by the quantum shot noise of light and quantum projection noise of atomic spins. Such sensitivity allows us to measure the nerve impulse with a miniature room-temperature sensor which is a critical advantage for biomedical applications. Positioning the sensor at a distance of a few millimeters from the nerve, corresponding to the distance between the skin and nerves in biological studies, we detect the magnetic field generated by an action potential of a frog sciatic nerve. From the magnetic field measurements we determine the activity of the nerve and the temporal shape of the nerve impulse. This work opens new ways towards implementing optical magnetometers as practical devices for medical diagnostics.

quant-ph↗

arXiv2015

Larger-Context Language Modelling

In this work, we propose a novel method to incorporate corpus-level discourse information into language modelling. We call this larger-context language model. We introduce a late fusion approach to a recurrent language model based on long short-term memory units (LSTM), which helps the LSTM unit keep intra-sentence dependencies and inter-sentence dependencies separate from each other. Through the evaluation on three corpora (IMDB, BBC, and PennTree Bank), we demon- strate that the proposed model improves perplexity significantly. In the experi- ments, we evaluate the proposed approach while varying the number of context sentences and observe that the proposed late fusion is superior to the usual way of incorporating additional inputs to the LSTM. By analyzing the trained larger- context language model, we discover that content words, including nouns, adjec- tives and verbs, benefit most from an increasing number of context sentences. This analysis suggests that larger-context language model improves the unconditional language model by capturing the theme of a document better and more easily.

cs.CL↗

arXiv2015

Automatic Instrument Recognition in Polyphonic Music Using Convolutional Neural Networks

Traditional methods to tackle many music information retrieval tasks typically follow a two-step architecture: feature engineering followed by a simple learning algorithm. In these "shallow" architectures, feature engineering and learning are typically disjoint and unrelated. Additionally, feature engineering is difficult, and typically depends on extensive domain expertise. In this paper, we present an application of convolutional neural networks for the task of automatic musical instrument identification. In this model, feature extraction and learning algorithms are trained together in an end-to-end fashion. We show that a convolutional neural network trained on raw audio can achieve performance surpassing traditional methods that rely on hand-crafted features.

cs.SD↗

arXiv2014

Is the Preferred Basis selected by the environment?

We show that in a quantum measurement, the preferred basis is determined by the interaction between the apparatus and the quantum system, instead of by the environment. This interaction entangles three degrees of freedom, one system degree of freedom we are interested in and preserved by the interaction, one system degree of freedom that carries the change due to the interaction, and the apparatus degree of freedom which is always ignored. Considering all three degrees of freedom the composite state only has one decomposition, and this guarantees that the apparatus would end up in the expected preferred basis of our daily experiences. We also point out some problems with the environment-induced super-selection (Einselection) solution to the preferred basis problem, and clarifies a common misunderstanding of environmental decoherence and the preferred basis problem.

quant-ph↗

arXiv2014

Strong micro-macro entanglement from a weak cross-Kerr nonlinearity

We study the entanglement generated by a weak cross-Kerr nonlinearity between two initial coherent states, one of which has an amplitude close to the single-photon level, while the other one is macroscopic. We show that strong micro-macro entanglement is possible for weak phase shifts by choosing the amplitude of the macroscopic beam sufficiently large. We analyze the effects of loss and discuss possible experimental demonstrations of the micro-macro entanglement based on homodyne tomography and on a new entanglement witness.

quant-ph↗

arXiv2014

Electronic Structures of Hybrid Graphene/Phosphorene Nanocomposite

Combining the electronic structures of two-dimensional monolayers in ultrathin hybrid nanocomposites is expected to display new properties beyond their simplex components. Here, first-principles calculations are performed to study the structural, electronic and optical properties of hybrid graphene and phosphorene nanocomposite. It turns out that weak van der Waals interactions dominate between graphene and phosphorene with their intrinsic electronic properties preserved. Hybrid graphene and phosphorene nanocomposite shows tunable band gaps at graphene's Dirac point and a transition from hole doing to electron doing for graphene as the interfacial distance decreases. Charge transfer between graphene to phosphorene induces interfacial electron-hole pairs in hybrid graphene and phosphorene nanocomposite with enhanced visible light response.

cond-mat.mtrl-sci↗

arXiv2014

First-Principles Study of Hybrid Graphene and MoS$_2$ Nanocomposites

Combining the electronic properties of graphene and molybdenum disulphide (MoS$_2$) monolayers in two-dimensional (2D) ultrathin hybrid nanocomposites have been synthesized experimentally to create excellent electronic, electrochemical, photovoltaic, photoresponsive and memory devices. Here, first-principles calculations are performed to investigate the electronic, electrical and optical properties in hybrid G/MoS$_2$ and G/MoS$_2$/G nanocomposites. It turns out that weak van der Waals interactions dominate between graphene and MoS$_2$ with their intrinsic electronic properties preserved. Interestingly, tunable p-type doping of graphene is very easy to achieve by applying electric fields perpendicular to hybrid G/MoS$_2$ and G/MoS$_2$/G nanocomposites, because electrons can easily transfer from the Dirac point of graphene to the conduction band of MoS$_2$ due to the work function of graphene close to the electronic affinity of MoS$_2$. Vertical electric fields can generate strong p-type but weak n-type doping of graphene, inducing electron-hole pairs in hybrid G/MoS$_2$/G sandwiched nanocomposites. Moreover, improved optical properties in hybrid G/MoS$_2$ and G/MoS$_2$/G nanocomposites are also expected with potential photovoltaic and photoresponsive applications.

cond-mat.mtrl-sci↗

arXiv2014

Proposal for the Creation and Optical Detection of Spin Cat States in Bose-Einstein Condensates

We propose a method to create "spin cat states", i.e. macroscopic superpositions of coherent spin states, in Bose-Einstein condensates using the Kerr nonlinearity due to atomic collisions. Based on a detailed study of atom loss, we conclude that cat sizes of hundreds of atoms should be realistic. The existence of the spin cat states can be demonstrated by optical readout. Our analysis also includes the effects of higher-order nonlinearities, atom number fluctuations, and limited readout efficiency.

quant-ph↗

arXiv2014

Analysis and Control of Beliefs in Social Networks

In this paper, we investigate the problem of how beliefs diffuse among members of social networks. We propose an information flow model (IFM) of belief that captures how interactions among members affect the diffusion and eventual convergence of a belief. The IFM model includes a generalized Markov Graph (GMG) model as a social network model, which reveals that the diffusion of beliefs depends heavily on two characteristics of the social network characteristics, namely degree centralities and clustering coefficients. We apply the IFM to both converged belief estimation and belief control strategy optimization. The model is compared with an IFM including the Barabasi-Albert model, and is evaluated via experiments with published real social network data.

cs.SI↗

arXiv2013

Can coarse measurements reveal macroscopic quantum effects?

It has recently been conjectured that detecting quantum effects such as superposition or entanglement for macroscopic systems always requires high measurement precision. Analyzing an apparent counter-example involving macroscopic coherent states and Kerr non-linearities, we find that while measurements with coarse outcomes can be sufficient, the phase control precision of the necessary non-linear operations has to increase with the size of the system. This suggests a refined conjecture that either the {\it outcome precision} or the {\it control precision} of the measurements has to increase with system size.

quant-ph↗